alpha-hunt

Phase 3 of the research program: divergent lever generation, then adversarial kill. Inputs: the three Phase 1–2 research files (all claims sourced there). This file is judgment, not new evidence — the main thread's synthesis.

The strategic center — the ownership fork

The market map's verdict reframes everything. Public evidence says GoProxies is a genuine third party buying Mysterium's bandwidth wholesale and retailing it at $1.40–2.59/GB (Lee's 80%-internal prior does not survive primary sources; Šaras's answer is pending and arbitrates — if he says "internal," the public surface misrepresenting their own structure becomes its own finding). Either way, the X-ray's first internal question is now sharp: what are the GoProxies deal economics — who captures the retail margin on Mysterium's own supply? From the answer, a fork:

• Branch A — own the retail layer: Mysterium-branded B2B storefront on its own network (or renegotiated capture from GoProxies). The Oxylabs mirror ($43.7M ARR, same city) shows what retail capture on this asset class looks like.
• Branch B — embrace wholesale, multiply storefronts: if retail is a knife-fight Mysterium can't staff (25 people, no sales hiring), do the opposite deliberately — replicate the GoProxies deal ×N with standard rev-share terms and become the neutral supply layer under many brands.

The fork is undecidable from outside. That's not a weakness — it IS the trial's week-one question, now precise instead of vague.

Divergence — 19 candidates generated

Revenue capture: (1) own retail storefront · (2) renegotiate GoProxies wholesale terms · (3) direct enterprise outbound arming the existing commercial director · (4) crypto-native self-serve access product (no-KYC, pay in stablecoin/MYST, API-first) · (5) publish list pricing + legal-entity/trust page · (6) premium-geo repricing. Cost side: (7) dVPN cost/automation audit (bigger team on the commodity product — break-even may be nearer via cost than revenue) · (8) sunset/merge consumer dVPN (mission-conflict, parked). Demand/GTM: (9) crypto-agent niche — agents that already hold wallets, natural no-KYC payers · (10) reseller-program multiplication (Branch B) · (11) narrative resurrection — founder voice + blog reboot (dead 3 years) · (12) Vilnius-cluster talent/partnership play · (13) "the web closes Sept 15" dated-catalyst campaign (Cloudflare default-block). Product: (14) FWA/mobile node-classification test (cheap, from prep) · (15) x402/MCP agent-payment prototype — agents autonomously buying bandwidth · (16) node-supply economics fix (thin lifetime payouts; supply not the bottleneck — parked). Partnerships/token: (17) Grass counter-position or partner · (18) DePIN-index visibility for crypto-native buyers · (19) revenue-linked token burn (only AFTER revenue moves — study verdict (b)).

The kill pass — criteria: 25-person-team-feasible, testable ≤90 days, doesn't fight funded incumbents head-on, moves break-even, SWAT-executable

Survivors (top 5):

• S1 · Resolve the ownership fork (1/2/10). Day-one internal question; both branches pre-designed so the answer converts to motion within days.
• S2 · Crypto-native self-serve wedge (4+9+15). The one uncontested space per the right-to-win analysis: protocol-level, no-KYC, crypto-paid access for decentralized and crypto-agent builders — a buyer no incumbent's compliance posture allows it to serve, native to Mysterium's actual architecture. Testable: thin storefront + outbound to ~20 named crypto-agent teams. The x402 prototype doubles as the demo AND the token's second-order setup.
• S3 · Trust-surface sprint (5). GoProxies' ToS names no legal entity; competitors lead with certificates. Publish list pricing, entity, trust page — near-zero cost, directly answers the found diligence gap, ships in week one regardless of the fork.
• S4 · Narrative reboot on a dated clock (11+13). Blog dead 3 years while the founder narrates live strategy; Sept 15 gives the campaign a countdown. Lee's native lane; feeds S2's pipeline and any branch of S1.
• S5 · dVPN cost audit (7). Internal-data dependent; the X-ray's cost half may show the shortest break-even path is cost, not revenue. Politically sensitive (mission product) — framed as "protect the mission product by making it affordable."

Killed — and why (the kill-list ships with the proposal):

• ✕ Generic AI-agent access outbound (the original T1): vendor-dense and freshly VC-funded (Browserbase $300M val, Firecrawl $15M A; Oxylabs/NetNut already ship agent SKUs). A 25-person team enters that knife-fight last.
• ✕ Head-on B2B proxy sales vs Bright Data/Oxylabs: suicide per right-to-win; Oxylabs alone is 397 people in the same city.
• ✕ Token-first revival: base rate ≈ low single digits from MYST's zone; every precedent needed a bigger trough + war chest + external wave. Demoted to second-order (19).
• ✕ "Ethical supply" as LEGAL differentiator: Bright Data's federal court wins (Meta, X) established scraping legality — ethics still matters for brand, not as legal moat.
• ✕ Consumer dVPN growth push: commodity race vs Nord in Nord's town.
• ✕ FWA/consented-mobile supply thesis: deflated in the 07-03 pass; survives only as the cheap classification test (14) inside S2's product work.

What this does to the proposal annexes

Annex A torpedoes rebuilt: T1' = crypto-native wedge probe (S2) · T2 = pricing probe (unchanged, now armed with GoProxies' public sheet) · T3' = trust-surface sprint (S3 — the competitor matrix already did the mystery-shop's desk half for free) · T4 = Šaras's test (parked, his definition). New annex: the outside-in X-ray summary + this kill-list — "we did the free half before day one."